SDPA extend backward realized tensor alignment checking to forward realized tensors #116069
Conversation
…alized tensors [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/116069
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 7160356 with merge base 87ea6fb ( UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
| def is_aligned(x): | ||
| return (V.graph.sizevars.size_hint(x.get_size()[-1]) % ALIGNMENT) == 0 | ||
|
|
||
| if isinstance(arg.data, ir.BaseView): |
There was a problem hiding this comment.
why doesnt is_aligned_realized_tensor work here?
There was a problem hiding this comment.
These inputs don't necessarily have strides()
|
@pytorchbot merge |
Merge failedReason: This PR needs a If not, please add the To add a label, you can comment to pytorchbot, for example For more information, see Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
…alized tensors (pytorch#116069) The logic to check alignment for realized tensors in the backward can be extended for realized tensors in the forward. This fixes an interaction with freezing. Pull Request resolved: pytorch#116069 Approved by: https://github.com/drisspg
3 participants
Stack from ghstack (oldest at bottom):
The logic to check alignment for realized tensors in the backward can be extended for realized tensors in the forward. This fixes an interaction with freezing.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler